Dealing with uncertainties related to ground motion prediction models for Georgia, Caucasus Region.

Georgia is situated in the Caucasus region, which is one of the most seismically active regions in the Alpine-Himalayan collision belt. Analysis of the historical and instrumental seismology of this region shows that it is still of moderate seismicity. The seismicity of the area reflects the general tectonics of the region. Recently, number of seismic stations and earthquake records in Georgia significantly increased. Thus, we can run more detailed studies regarding ground motion prediction.  Ground motion prediction equations (GMPEs) relate ground motion intensity measures to variables describing earthquake source, path, and site effects. In this study ground motion prediction equations are obtained by classical, statistical way, regression analysis. Also, new data and new features such as local soil conditions, fault types, etc. were considered for analysis. In the study models are obtained for PGA (horizontal and vertical), 5%-damped pseudo-absolute-acceleration spectra (SA) are described for periods between 0.01 s and 10 s (for both vertical and horizontal components). Next stage was to assess the standard deviation and its minimization. Fuzzy Analysis gives a possibility of making optimal decision when available data is insufficient and cannot represent real situation. In our case it is quite difficult to explain all physical processes related to earthquakes. However, it is very important to consider all processes during the hazard assessment. Also, during GMPE assessment it is very difficult to consider site effect very precisely because available data is still insufficient. In this case usage of Fuzzy Analysis is the best solution. We constructed membership functions based on shear wave velocity measurements for each site class. Site classifications were done according to Eurocode8. At the end a significant reduction of uncertainties (~30-40%) was observed.